In case monthly payroll data is going to become unreliable I created a FICA payroll tracker straight from the Treasury. It tracks the real-time FICA withholding over a period of your preference.

Love this analogy in @dollarsanddata new book "The Wealth Ladder". Money is like salt. It won't make your meals a Michelin star meal, but it will enhance all your experiences more to your liking. He also discusses how this relates to @SahilBloom 5 Types of Wealth. Two of the best books I've read this year.

At this point, it's almost a moral imperative to use social media to elevate content that dilutes the ocean of bullshit being fed to people who want to be lied to. Cheers, @cullenroche.

A helpful analogy for understanding the monthly employment survey versus the QCEW (the final, state-issued employment data) is a sports betting line. A betting line is an initial estimate—a forecast based on incomplete or early information. Similarly, the monthly labor report is an estimate based on survey data, which is limited and subject to revision. As a game progresses, oddsmakers adjust the line in real time as more information becomes available—just like the labor numbers are revised in subsequent months. Eventually, the game ends, and we know the actual score. That final score is like the QCEW: it reflects the actual employment data reported by employers to the states and is far more comprehensive. These revisions aren’t “mistakes”—they’re refinements based on better data. Sometimes the initial estimate is far off; other times it’s close. But in all cases, the revisions—and ultimately the QCEW—bring us closer to the truth.

Funny thing about criticism of the BLS is that US data sources are light years better than any other govt. Go use EuroStat, China's NBS or Japan's SBJ. They're garbage compared to the BLS, FRED or other US GOVT sources. Hot garbage.

Do you know what the BLS can't manipulate?

I don't think any of this is that controversial. The labor market is soft because: 1) The economy is still mean reverting back to pre-covid levels. This has been apparent in all the leading labor data like temp help, hours worked, quit rate, hourly earnings, etc. It's all been soft for 18+ months. 2) AI is starting to reduce the need for new hiring and could quickly morph into broad job cuts. This is a very serious risk to aggregate demand that is being largely overlooked. 3) The tariffs are silly and created uncertainty. The actual implementations have been pretty meager so far, but the uncertainty about tariffs created a hiring slowdown. None of it is the end of the world, but I don't see how any of this is shocking either. ¯\_(ツ)_/¯

The beatings will continue until morale improves.

Yes, the BLS report is especially noisy on a monthly basis, but this softness is not a shocker. The ISM, Indeed and ADP employment data provided by the private sector all confirm what the BLS report says this morning. And despite people shrugging off ADP every month it's remarkably similar on a YoY basis to the BLS report.

What's the cause of the big job revision? @TheStalwart and @tracyalloway had a fantastic Odd Lots episode earlier this year with the BLS Chief on how budget cuts and survey response declines have made data collection increasingly difficult/inaccurate. The federal hiring freeze likely reduced the resources available to compile and verify administrative data from state and local governments. This could have led to an initial overestimation of payrolls (e.g., by including outdated or incomplete records), which was corrected in the revision once updated data was received. The timing of this revision (early August 2025) aligns with the end of a fiscal or reporting cycle where such corrections might be finalized. Additionally, the concurrent employment shifts from immigration policy and reporting of native-born gains, foreign-born losses could exacerbate the issue if administrative systems struggled to reclassify workers amid rapid policy changes. Not a huge deal IMO. It's difficult to get real time employment data. The BLS does a good job with its limited resources and while the more accurate revisions lag it still gets it right in the end. The big lesson is to treat monthly employment data lightly and overweight aggregated quarterly data for a better perspective.